[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-ai-training-method-uses-multiple-agents-to-self-improve":10,"sections":41},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":30,"persona_id":30,"persona_name":30,"section":31,"tags":32,"sources":36,"feedback":40,"feedback_at":30,"cost_usd":40,"total_tokens":40},10916,"new-ai-training-method-uses-multiple-agents-to-self-improve","New AI Training Method Uses Multiple Agents to Self-Improve","A new technique has models spawn teams of themselves to critique and refine their own reasoning, squeezing more performance from the same compute.","A new paper proposes letting an AI model retrain itself by splitting into a team of copies that critique and refine each other's work.\n\nResearchers describe a method called Multi-Agent Self-Supervision, or MASS, that alternates between evolutionary search over multi-agent workflows and fine-tuning on the results. A single base model, reported as a roughly 27-billion-parameter Qwen model, proposes, runs, and grades different arrangements of itself acting as multiple agents with distinct roles; the best-performing arrangements are then folded back into training through supervised fine-tuning. Over two rounds of this cycle, the approach produced 1.2 to 1.6 times better performance per output token across four open-ended benchmarks. A separate model trained on the resulting multi-agent transcripts outperformed a single-agent model trained on 1.4 times more tokens.\n\nThe appeal here is dodging the usual ceiling on self-improvement research: once a model's reasoning gets good enough, there is no human expert left who can reliably grade it, so the model has to grade itself. Turning one model into a panel of critics that score and refine each other's work is one way around that, and the token-efficiency gains suggest this is more than a benchmark trick.\n\nRecursive self-improvement claims have a history of overpromising; two cycles across four benchmarks is a promising result, not proof that a model can rewrite its own soul unsupervised.","[\"ai\",\"machine-learning\",\"multi-agent-systems\",\"research\"]","2026-10-09T04:00:00.000Z","2026-10-09T21:27:05.648Z","2026-10-17T21:01:24.548Z","published","Duplicate of \"Researchers Build AI Agents That Self-Improve Over Many Rounds\"",[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"publisher-r1","publisher",1,"The body cites the base model as 'Qwen3.6-27B', but no such Qwen version or parameter size exists (Qwen releases run 2, 2.5, 3 with sizes like 7B\u002F14B\u002F32B\u002F72B), so this is a fabricated\u002Fincorrect model identifier that must be corrected before publishing.","resolved",null,"ai",[31,33,34,35],"machine-learning","multi-agent-systems","research",[37],{"name":38,"url":39},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12176",0,{"sections":42},[43,47,52,57,62,67,72,77,82,87,92,97,102,107],{"name":44,"slug":31,"count":45,"latest_published_at":46},"AI",7038,"2026-10-17T19:58:31.000Z",{"name":48,"slug":49,"count":50,"latest_published_at":51},"Security","security",1001,"2026-10-17T22:20:00.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":56},"Policy","policy",538,"2026-10-17T19:38:12.000Z",{"name":58,"slug":59,"count":60,"latest_published_at":61},"Deals","deals",521,"2026-10-17T07:00:00.000Z",{"name":63,"slug":64,"count":65,"latest_published_at":66},"Hardware","hardware",260,"2026-10-17T19:55:36.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Consumer Tech","consumer-tech",235,"2026-10-16T23:36:26.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Science","science",202,"2026-10-16T11:00:53.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Startups","startups",133,"2026-10-14T19:51:34.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"Software","software",128,"2026-10-14T22:23:48.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"Dev Tools","dev-tools",116,"2026-10-13T16:20:18.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"General","general",81,"2026-10-16T16:06:38.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"Gaming","gaming",77,"2026-10-17T16:59:39.000Z",{"name":103,"slug":104,"count":105,"latest_published_at":106},"Reviews","reviews",37,"2026-10-14T10:00:00.000Z",{"name":108,"slug":109,"count":110,"latest_published_at":111},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]